LIVER transplantation remains the gold standard for treating patients with early-stage hepatocellular carcinoma, yet the scarcity of donor organs necessitates precise patient selection.
Traditionally, criteria like the Milan guidelines have been used to assess eligibility, but newer, more nuanced models have emerged. Among them is the TRIUMPH model, a machine learning-based risk prediction tool that assesses the likelihood of post-transplant recurrence. In this international study, researchers externally validated TRIUMPH using a large multicentre cohort and compared its performance with existing models such as AFP, MORAL, and HALT-HCC.
The findings revealed that the TRIUMPH model offered numerically superior accuracy in predicting recurrence, particularly in subgroups involving living donor liver transplantation and non-US transplant centres. Its strength lies in incorporating a wider array of pre-transplant factors, including tumour morphology, biomarkers like alpha-fetoprotein, and patient responses to bridging therapies. These improvements led to a higher net clinical benefit in decision-making analyses, especially across realistic transplant probability thresholds.
While other machine learning models like MoRAL-AI and TRAIN-AI have also shown promise, TRIUMPH distinguishes itself through its robust generalisability and its focus on pre-operative data, which is critical for organ allocation decisions. Nevertheless, differences in regional transplant practices and patient populations affected model performance across centres, underlining the importance of international collaboration in future development.
Despite being retrospective and multicentric, the study supports TRIUMPH as a valuable predictive tool that reflects real-world transplant scenarios, involving both living and deceased donors. The model’s integration into clinical practice could lead to more equitable and effective use of donor organs. Still, widespread adoption will require further validation, logistical planning, and consensus among the global transplant community. Overall, TRIUMPH represents a step forward in applying machine learning to complex medical decision-making, offering a potentially transformative approach to organ allocation in hepatocellular carcinoma.
Reference
Li Z et al. Validation of the Toronto recurrence inference using machine-learning for post-transplant hepatocellular carcinoma model. Commun Med (Lond). 2025;5(1):284.
